Neurofeedback Beyond EEG: A Systems-Level View

1. Intro and brief history

Part 4 of this series stepped us out of Layer 1 — the autonomic substrate that neurofeedback protocols are trained inside of — and into Layer 3, external modulation. taVNS was the entry point into our active neuromodulation methods, and it earned the slot. This fifth piece stays in Layer 3 but turns to a modality I treat differently. If taVNS is the Layer 3 tool I am building an integration framework for, photobiomodulation is the Layer 3 tool I am waiting on. Both deserve a careful, full-length look. Taking a modality seriously does not always mean adopting it — sometimes it means explaining, precisely and fairly, why the case for adoption has not yet been made.

A quick orientation for any reader picking up the series here. As you probably know, neurofeedback is a form of biofeedback in which a person receives real-time feedback about their own brain activity (usually EEG) and learns, through operant conditioning and voluntary self-regulation, to shift that activity in a desired direction. Biofeedback is the broader family — real-time feedback about a physiological signal (heart rate variability, skin conductance, muscle tension, temperature) used to train self-regulation of that signal. Both are learning interventions: the system is taught to act on itself. Photobiomodulation is not a learning intervention at all. It acts on the system from outside — it is active modulation — which is exactly what puts it in Layer 3, and exactly what raises the methodological and ethical bar relative to the substrate tools in Layers 1 and 2.

So what is it? Photobiomodulation is the therapeutic application of red (roughly 600–700 nm) and near-infrared (roughly 780–1100 nm) light to biological tissue, at intensities low enough that the primary effect is photochemical and photobiological rather than thermal. You shine light of a particular wavelength, power density, and dose at tissue, and — the claim goes — you change what the cells under that light are doing, principally by way of their mitochondria. When the target is the brain and the light is delivered through the scalp, we call it transcranial PBM (tPBM), and that is the application relevant to this series.

The history is somewhat old, which is part of why the field's confidence sometimes outruns its evidence. In 1967, Endre Mester in Budapest, working with a low-powered ruby laser and trying (he thought) to test whether laser light could cause cancer in mice, noticed instead that shaved fur grew back faster on the irradiated animals. That accidental observation launched what became known as low-level laser therapy. For the next four decades, PBM lived mostly in dermatology, dentistry, sports medicine, and wound care, where the peripheral evidence base for soft-tissue healing, pain, and inflammation accumulated into something reasonably solid. The brain came late to the party, but the transition seems only natural. Transcranial PBM as a serious research program is largely a phenomenon of the 2000s and 2010s — riding on improvements in LED and laser sources, on the mitochondrial-mechanism work, and (let us be honest) on a wellness market that discovered red light could be sold by the panel, the helmet, and the intranasal clip.

The honest framing in 2026: PBM is an old peripheral modality with a real cellular mechanism and a genuinely promising — but young, heterogeneous, and dosimetrically fraught — transcranial research literature. What will you read in the slogans? Unfortunately, some are already positioning this as either miracle light helmet or expensive placebo, and it is the position I want to help you see between. It is also, I will say up front, a more skeptical position than I took on taVNS, and the skepticism is earned rather than reflexive. I don’t think it’s an expensive placebo, but I’m watching and waiting to see what the science says.

Science is slow, clinical practice often goes fast… but I personally have learned to trust a healthy balance of the two. I want to be the person who does the right thing at the right time, not the one who does the new thing first.

2. Alternate names

The terminology around therapeutic light has shifted over time, and — as with vagal stimulation — the names carry baggage worth a paragraph.

Low-level laser therapy (LLLT) is the historical term, and you will still see it everywhere in the older literature. It reflects the field's laser-centric origins and the "low-level" distinction from surgical or ablative lasers that cut and coagulate. The problem with the name is that it baked in two assumptions that turned out to be wrong or limiting: that the source had to be a laser (LEDs work too, and are cheaper and safer), and that "low-level" was a meaningful dosing category (it is not — dose is a continuous variable with a biphasic response, which I will get to).

Photobiomodulation (PBM) is the term the field consolidated around in the mid-2010s, formally adopted by the North American Association for Photobiomodulation Therapy and the World Association for Photobiomodulation Therapy and added to the medical subject headings around 2015. It is the better term precisely because it is mechanism-agnostic about the source (laser or LED), and because modulation honestly captures that the effect is a nudge to cellular function rather than a fixed pharmacological action. Photobiomodulation therapy (PBMT) is the same thing with the clinical-application suffix.

Transcranial photobiomodulation (tPBM), transcranial low-level laser/light therapy (tLLLT), and transcranial near-infrared light therapy all name the brain-directed application specifically. Transcranial infrared laser stimulation (TILS) shows up in the cognitive-neuroscience literature, particularly the University of Texas work on prefrontal stimulation and cognition. Near-infrared stimulation (NIRS) is unfortunately collision-prone — the same acronym is near-infrared spectroscopy, a measurement technique — so context matters.

Then there is the consumer and wellness vocabulary, which is its own ecosystem: red light therapy, infrared therapy, cold laser, low-level light therapy, photonic therapy, and a long tail of branded device categories. Vielight (intranasal plus transcranial LED clusters), various transcranial helmets and caps studded with LED arrays, and full-body or panel red light systems all live here. Quality, wavelength accuracy, irradiance, dose control, and — critically — the actual fraction of light delivered to the intended target vary enormously across this space, and the regulatory boundary between medical device and wellness wearable is even more fluid than it is for EEG or for the vagal stimulators in the last piece.

The information you actually need behind any of these names is the same short list every time: what wavelength, at what irradiance (power density, mW/cm²), for what dose (fluence, J/cm²), delivered to what target, with what evidence that the light actually arrives there. A 1064 nm laser delivering a measured fluence to prefrontal cortex in a controlled cognitive-neuroscience experiment is one intervention. A consumer LED helmet of unspecified per-diode irradiance worn for twenty minutes is a very different intervention that happens to share a name. They are not the same thing therapeutically, and the name does almost none of the work of telling them apart.

3. How it works

The proposed mechanism of PBM is, at the cellular level, more concrete and more biologically grounded than the marketing's vagueness would lead you to expect — which is part of what makes the modality frustrating. There is a real story here. The trouble starts when you try to connect the real cellular story to a real clinical effect through a centimeter of scattering bone.

Start with the light. Red and near-infrared wavelengths occupy what is sometimes called the optical window of biological tissue — roughly 650–1200 nm — the band in which light penetrates tissue relatively best because it is absorbed least by the major chromophores that otherwise eat photons. Below this window, hemoglobin and melanin absorb strongly; above it, water absorbs strongly. In the optical window, light still scatters heavily, but it is absorbed less, so it travels further before being extinguished. This is why PBM uses red and NIR specifically, and why the choice of wavelength within the window is not cosmetic — penetration depth and the absorption profile of the target chromophore both depend on it.

The primary photoacceptor — the molecule that absorbs the light and starts the biological cascade — is generally held to be cytochrome c oxidase (CCO), also called Complex IV, the terminal enzyme of the mitochondrial electron transport chain. This is the core of the mechanistic hypothesis, developed substantially through the work of Tiina Karu and elaborated by Michael Hamblin and others. CCO has absorption peaks in the red and near-infrared, and the leading model holds that photon absorption by CCO causes the photodissociation of nitric oxide from the enzyme. Nitric oxide, under metabolic stress, inhibits CCO by competing with oxygen at the binding site; knock the NO off with a photon, and you relieve the inhibition, restore electron transport, and increase the proton gradient and downstream ATP production. The light, on this account, does not add energy to the cell in any meaningful caloric sense — it removes a brake.

From that proximate event, the proposed downstream cascade fans out: a transient, low-level rise in reactive oxygen species acting as a signaling molecule (not as oxidative damage, at appropriate doses); release of nitric oxide into the local environment, with vasodilatory and blood-flow consequences; activation of transcription factors (NF-κB and others); and a longer arc of effects on cerebral blood flow, mitochondrial biogenesis, neurotrophic signaling (BDNF), inflammation, and — in the brain-specific literature — proposed effects on amyloid handling and neuronal metabolism. There are also secondary mechanistic candidates that are less settled: light-sensitive ion channels of the TRP family, effects on interfacial water structuring near membranes, and direct effects on the heat-shock response. The honest state of the science is that CCO photodissociation of NO is the leading hypothesis with real supporting data, and that the full pathway from photon to durable clinical change is not closed.

The procedure, in practice, varies enormously, and that variation is a problem rather than a feature. A researcher running a controlled tPBM cognition study might use an 810 nm or 1064 nm laser at a carefully measured irradiance, targeting a specific scalp location over prefrontal cortex, for a defined number of minutes calibrated to deliver a specific fluence. A clinical TBI or depression protocol might use LED clusters at one or two wavelengths (commonly 810 nm, sometimes paired with 633 nm red), applied to multiple scalp sites and often an intranasal emitter, several times per week for weeks. A consumer red-light panel might bathe the whole front of the head in mixed-wavelength light of unspecified per-area irradiance for whatever duration the user chooses. These are not dosing variations on a theme; they are different interventions with wildly different odds of delivering a biologically meaningful dose to any intended target.

Which brings us to the parameter that quietly governs everything: dose, expressed as fluence in joules per square centimeter, and its rate, irradiance in milliwatts per square centimeter. PBM follows a biphasic dose-response — the Arndt-Schulz curve, characterized in this context by Ying-Ying Huang, Hamblin, and colleagues. Too little light does nothing. The right dose produces the beneficial photobiological effect. Too much light overshoots into inhibition or null effect — more is not better, and past a point more is worse. This matters more than almost anything else in the modality, and it is the parameter the consumer market handles worst, because getting the dose at the target right requires knowing how much light started, how much was lost on the way, and how big the target is. For peripheral applications — light on skin, a centimeter from the muscle — that calculation is hard but tractable. For the brain, it runs straight into the wall.

4. Mechanistic specifics

Three things are worth naming explicitly, because each grounds a different part of the clinical assessment and the eventual integration question.

The first is the chromophore and the energy logic. The CCO-photodissociation model is attractive partly because it is parsimonious and partly because it makes the right kind of prediction: a brake-release on mitochondrial respiration would plausibly produce broad, non-specific, metabolism-supporting effects across many tissues and conditions — which is exactly the clinical signature PBM shows. That breadth is a double-edged sword. It is mechanistically coherent, and it is also the reason PBM gets marketed for essentially everything, because a mitochondrial-support story can be told about almost any condition. A mechanism that explains everything explains nothing in particular, and the clinician's job is to demand that the general cellular story be cashed out into a specific, dosed, target-verified, outcome-measured claim before it counts for anything in a treatment plan.

The second — and this is the important bit — is the penetration and dosimetry problem. For transcranial PBM to do anything at all in the brain, photons have to traverse, in order: hair (which absorbs and scatters, and varies by color and density — take, for example, those more aerodynamically predisposed such as myself [read: bald]), scalp (vascular, absorbing), the skull (which both absorbs and, crucially, scatters light strongly), the meninges, and cerebrospinal fluid — before reaching cortex, and far more before reaching anything subcortical. At each interface and through each layer, light is attenuated. Cadaver-tissue and Monte Carlo modeling studies that have quantified this transmission — the Jagdeo and colleagues cadaver work, the Tedford and colleagues human-brain-tissue penetration measurements, and several modeling papers since — converge on a sobering picture: only a small fraction of incident light reaches the cortical surface, and the fraction reaching depth (a few centimeters in, where much of the interesting brain lives) is smaller still. The exact numbers depend on wavelength (longer NIR penetrates better than red), on whether the source is collimated, on skull thickness, and on the measurement method, so I will not pin a single figure to it — but the order of magnitude is the point. The light that reaches the cortex is a slim minority of the light that left the device, and for whole-head consumer systems whose per-area irradiance is modest to begin with, the delivered cortical dose can fall to a small fraction of one percent of what was emitted. I covered this in an earlier NeuroBLOG, and the headline holds up: for some consumer transcranial systems, well over 99% of the light does not make it to the brain. That is not a reason PBM cannot work — the research systems are engineered around exactly this problem, with higher irradiance, better wavelength selection, and sometimes laser sources — but it is the reason a price tag and a glowing helmet tell you almost nothing about whether a meaningful dose arrived where it was supposed to.

The third is the pulsing and timing question, which I raise mostly to flag how unsettled the parameter space is. Continuous-wave versus pulsed delivery, pulse frequency, and the interaction of pulse parameters with the biphasic dose curve are all actively debated, with some evidence that pulsed delivery at particular frequencies behaves differently from continuous-wave at matched average power. The relevant point for a practitioner is not which side is right; it is that a modality whose basic delivery parameters are still under genuine scientific dispute is not a modality whose clinical protocols can be considered settled. When the field cannot yet agree on whether to pulse the light and at what frequency, the clinic should not be pretending the recipe is known.

A distinction I want to lock down, the same way I did for taVNS: the cellular mechanism is well-supported; the peripheral clinical effects are reasonably established; the transcranial clinical effects are early and heterogeneous; and the delivered-dose-to-target question sits underneath all of it as an unsolved engineering and measurement problem for the consumer tier. Each of those four claims has a different evidentiary weight, and the most common error in the marketing — and in some of the enthusiast clinical practice — is to let the strength of the first claim leap forward to underwrite the fourth.

5. Overview of the science base

The evidence base for PBM is large, moderately old, and extremely uneven across applications. A fair summary has to separate the periphery from the brain, and within the brain, further divide the mechanistic-cognitive work from the clinical-outcomes work.

Peripheral PBM — light applied to skin, joints, muscles, and wounds — is the part of the field with the most credible track record. Across musculoskeletal pain, tendinopathy, oral mucositis (where it is genuinely a standard of supportive care in some oncology settings), and wound healing, there are systematic reviews and meta-analyses supporting real, if modality-specific and dose-dependent, effects. I mention this because it matters for fairness: a clinician who dismisses PBM wholesale is overreaching in the other direction. The cellular mechanism is real, and where the target is a centimeter under the skin, the dosimetry is tractable and the outcomes can be respectable. The periphery is not the problem. The brain is.

Transcranial mechanistic and cognitive work in healthy adults is where some of the cleaner experimental signal lives. The Gonzalez-Lima and Barrett line of work at the University of Texas, among others, has reported acute effects of prefrontal tPBM (often 1064 nm laser) on measures of attention, working memory, and prefrontal oxygenation in healthy participants, with some sham-controlled designs. These are real experiments with measured doses and plausible effects, and they are the strongest argument that transcranial light can do something measurable to the cortex it reaches. They are also small, acute, mostly in healthy young adults, and a long way from a durable clinical indication.

Transcranial clinical outcomes are where the picture gets genuinely thin and heterogeneous. In traumatic brain injury, the Naeser and colleagues case series and small studies reported cognitive and functional improvements with LED tPBM, but the work is mostly open-label, small, and uncontrolled — proof-of-concept, not efficacy. In depression, there have been small sham-controlled pilots — the Schiffer early work, and the Cassano and colleagues transcranial PBM pilot in major depressive disorder — with signals that are encouraging at pilot scale and nowhere near the evidentiary level that would move a treatment guideline. In cognitive aging and dementia, there is early, mostly preliminary work. In Parkinson's disease, transcranial and systemic PBM has an active and enthusiastic research community and a genuinely interesting mechanistic rationale, but the clinical evidence remains early-stage. Across all of these, the pattern is the same: promising mechanism, small studies, open-label or small sham-controlled designs, heterogeneous parameters, and a wide gap between the marketing's confidence and the literature's.

And then there is the cautionary tale the field does not advertise. The NEST program — the NeuroThera Effectiveness and Safety Trials — tested transcranial near-infrared laser therapy for acute ischemic stroke across three trials in the 2000s and early 2010s. NEST-1 was a promising small trial. NEST-2 was mixed, with a possible signal in a subgroup. NEST-3, the large pivotal phase III trial, was stopped early for futility. This is the most rigorously tested transcranial PBM clinical program to date, and it failed at the pivotal stage. There are reasonable mechanistic debriefs — wrong dose, wrong timing, the penetration problem, stroke being a hard target — and the failure of one attempt to establish an indication does not condemn the modality (not even within the specific domain tested; replication is still needed). But it is exactly the kind of result that should temper anyone's confidence, and it is conspicuously absent from the consumer-device sales pitch. A modality whose single largest, best-controlled brain trial ended in futility is a modality to watch carefully, not one to build a practice on.

The honest summary: peripheral PBM is reasonably established and dose-tractable; transcranial PBM has a coherent cellular mechanism, a thin and heterogeneous clinical literature, one prominent pivotal failure, and an unresolved dosimetry problem that the consumer market largely ignores. The next decade of well-engineered, properly dosed, sham-controlled trials will decide whether transcranial PBM becomes a defined clinical tool or stays a mechanistically interesting frontier. My bet is that it earns a place for a subset of indications under research-grade delivery — and that most of what is currently sold to consumers as brain photobiomodulation will not be the thing that gets there.

6. Strengths and weaknesses

What does PBM have going for it, and where does it fall down?

Strengths

The cellular mechanism is real and reasonably well-characterized — this is not a modality built on hand-waving, and the CCO story has decades of basic-science support behind it. The safety profile is excellent at typical doses: non-invasive, painless, no meaningful systemic risk, with the main genuine hazards being thermal (at high irradiance) and ocular (with laser sources, which is why eye protection is non-negotiable in laser protocols). The peripheral evidence base for pain and wound healing is respectable and gives the modality a legitimate clinical foothold. And the brain-directed mechanistic work in healthy adults shows that transcranial light can produce measurable cortical and cognitive effects under controlled, well-dosed conditions — which is a real foundation to build on, even if the building is not done. PBM is also, unlike rTMS, operationally light: no session-tying equipment footprint, no motor-threshold determination, no specialist-clinic infrastructure. If it worked reliably at the consumer tier, it would be unusually easy to integrate. That conditional is doing a lot of work.

Weaknesses

The dosimetry problem is the headline, and it is not a detail — it is the central unsolved issue for transcranial application, and the consumer market's near-total silence about it is itself a red flag. The transcranial clinical evidence is early, small, mostly open-label or small-pilot, and heterogeneous to the point where cross-study comparison is barely possible: wavelength, irradiance, fluence, pulsing, site, session number, and target indication all vary, and the biphasic dose-response means that two studies using "the same" device at different durations can be testing genuinely different interventions. The single best-controlled pivotal brain trial (NEST-3) failed. The mechanism's very breadth — mitochondrial support for everything — makes it marketable for everything, which is precisely the property that should make a careful clinician suspicious rather than reassured. And the gap between research-grade delivery (measured irradiance, validated wavelength, sometimes laser sources, careful dosing) and consumer-grade delivery (a helmet of LEDs at unspecified per-area irradiance) is enormous, expensive to close, and almost never disclosed at the point of sale.

A common failure I want to name directly, because it is the one this series keeps circling: importing the peripheral or the mechanistic evidence wholesale into a transcranial clinical claim. "PBM heals wounds and reduces inflammation, and inflammation is involved in depression, therefore our light helmet treats depression" is a syllogism doing far more work than the evidence supports — and it skips the entire question of whether a therapeutic dose reached the relevant tissue. "Light activates mitochondria, your brain is full of mitochondria, therefore this device optimizes your brain" is the same move in consumer clothing. The general cellular truth does not underwrite the specific clinical claim, and the dosimetry gap sits in between them unacknowledged.

A second failure is under-specifying the intervention, the same discipline problem the whole series keeps returning to. "We do some red light therapy" is not an intervention description. 810 nm LED at a measured irradiance of X mW/cm², delivered to defined scalp sites for Y minutes to achieve a target fluence of Z J/cm², three times weekly for six weeks is — and notice that even that description is incomplete without an estimate of the fraction actually reaching the target. The field has had to learn parameter discipline for EEG-NF; PBM needs it more, not less, because its biphasic dose-response means the wrong dose is not merely weaker but potentially inert or counterproductive.

A third is confusing the wellness-market enthusiasm for clinical validation. The volume of red-light devices, testimonials, and confident YouTube mechanism explainers is not evidence. It is a market. The clinical literature is the evidence, and the clinical literature, for transcranial brain targets, says promising and not yet there.

7. Brendan's perspective

The single idea I most want you to remember from this post: I would not currently reach for photobiomodulation in my neurofeedback practice, and I think the reasons are worth laying out carefully — because the case against adopting it now is also the spec sheet for what would make me change my mind.

Let me say plainly why it is a no, for now. There are four reasons, and they stack.

The first is that it is too young. Not the modality — the modality is older than I am. The transcranial brain-clinical evidence. What exists is mostly small, open-label, heterogeneous proof-of-concept work, plus a clean-ish acute cognitive-neuroscience literature in healthy adults, plus one large pivotal trial in stroke that was stopped for futility. That is not the evidence profile of a tool I integrate into a clinical workflow where I am accountable for outcomes. It is the evidence profile of a tool I read about and keep a folder on. I held taVNS to this same bar and it passed for a narrow set of cases; PBM does not pass it yet for any case I actually see.

The second is that the mechanism, at the dose that matters, is unclear. I want to be fair here, because the cellular mechanism is genuinely well-supported — cytochrome c oxidase, nitric oxide photodissociation, the metabolic brake-release. That part I believe. What I do not yet believe we understand is the chain from that cellular event to a durable clinical change in a human brain at the dose that actually arrives through a living skull. The marketing tells a confident end-to-end story. The science supports the first link and gets progressively shakier toward the last, and the breadth of the mechanism — mitochondria support everything, so light helps everything — is exactly the kind of explanation that should raise a clinician's eyebrow rather than lower their guard.

The third is the one I have written about before and will keep writing about: the skull. For transcranial PBM to work, light has to get to the brain, and the scalp and skull scatter and absorb most of it before it arrives. The careful penetration work — cadaver studies, tissue measurements, Monte Carlo models — keeps landing on the same uncomfortable order of magnitude: a small fraction of incident light reaches cortex, and for whole-head consumer systems with modest per-area irradiance, the delivered cortical dose can fall below one percent of what was emitted. I wrote a whole NeuroBLOG on how, for some consumer systems, north of 99% of the light never reaches the brain at all. The research-grade systems are engineered specifically to fight this — more irradiance, better wavelengths, sometimes lasers, careful dosimetry — which is precisely why you cannot read across from "a trial showed an effect" to "this glowing helmet will reproduce it." (They do tend to look really cool, though.) The trial's dose and the helmet's dose are not the same dose, and with a biphasic response curve, dose is the whole game.

The fourth reason is the least scientific and the most practical, and I am going to be honest about it: it is expensive. To do transcranial PBM in a way that has any business being called clinical — research-validated wavelengths, measured irradiance, enough delivered dose to plausibly matter, the kind of engineering the credible trials used — costs real money. And the cheaper you go, the more you slide toward the consumer tier where the dosimetry falls apart and you are, functionally, selling people a head-worn version of those neon lights you sometimes see under a souped-up coupe. So the choice is between an expensive system whose clinical evidence is still too thin to justify the outlay, and an affordable system whose physics make the clinical claim implausible. Neither side of that trade is one I want to make right now, and I would rather tell a client that honestly than sell them either version.

Now the part that makes this a perspective and not just a dismissal. What would change my mind? Concretely: well-engineered, properly dosed, sham-controlled transcranial trials in indications I actively work with, with delivered-dose verification rather than emitted-dose assumptions, replicated beyond a single enthusiastic lab. A consumer or clinic device tier that publishes its per-area irradiance and an honest estimate of delivered cortical dose, so the dosimetry stops being a black box. And ideally, some direct work on the question this series cares about most — whether transcranial PBM, applied around a neurofeedback session, does anything to the learning. That last question is genuinely interesting, and I do not want my skepticism about the current state to be heard as a claim that the idea is foolish. The idea is fine. The evidence and the engineering are not there yet.

The brand-voice tripwire I want to be careful around — and it is the same one from the first landscape piece of the series — is the cash grab of the overeager innovator. Our field attracts open-minded, curious clinicians; that openness is a strength and it is also the exact psychological surface that aggressive light-device marketing is engineered to press on. Open-minded is not the same as open season. I can hold genuine interest in transcranial PBM's future and a firm "not in my practice yet" at the same time, and I think that is the honest posture. It is not nothing, and it is not magic either — and right now, for the brain, through the skull, at consumer prices, it is closer to not yet than to either.

TL;DR PBM has a real cellular mechanism and a respectable peripheral track record, but transcranial brain use is too young, the mechanism-at-delivered-dose is unclear, most of the light never reaches the brain, and the version that might work is expensive. I am watching it closely and not integrating it — and I have written down exactly what would change that.

8. Would I integrate PBM into my NF practice? In what context?

(If you skipped ahead to this part: the short answer is no, not yet — and the rest of this section is the careful version of why, plus the conditions under which the answer would flip.)

The integration question for PBM is the mirror image of the one for taVNS, and the contrast is the whole point of putting these two Layer 3 pieces near each other. taVNS got a yes, selectively — because it engages the autonomic and attentional axis neurofeedback already trains, the mechanism is clean, the dose actually reaches its target, and the acute effects are real. PBM gets a not yet — because the transcranial clinical evidence is too young, the mechanism-to-clinical-outcome chain is unproven at the dose that arrives, most of the light does not arrive at all, and closing that gap is expensive. Same series, same framework, opposite verdict. That is not inconsistency; it is the framework doing its job.

Let me give the honest clinical answer in the same four-layer shape I used for taVNS — who, when, how, and how I would talk about it — except here most of the layers resolve to not currently, and here is what would have to be true.

Who I would consider it for — under what conditions. Today: essentially no one in my routine practice, because the conditions are not met. In a future where the evidence matured, the candidate would be a client with a specific, defensible metabolic or perfusion rationale — not a generic "boost your brain" rationale — in whom the substrate work and the neurofeedback had been worked through, and for whom a research-validated delivery system with verifiable dosimetry was available. The specificity of the rationale is the gate. "This client is hypoaroused, let us add light" is not a rationale; it is a vibe. "This client has a specific presentation for which dosed transcranial PBM has replicated, sham-controlled evidence and a plausible delivered dose" would be — and that sentence is currently hypothetical.

When in the treatment arc I would introduce it. Never as a substrate tool and never as a substitute for the learning work, because PBM does not teach the system to regulate itself — it acts on the system from outside. If it earned a role, it would be late and adjunctive: after the Layer 1 substrate tools and the neurofeedback itself had been deployed and assessed, in the same slot where I think about other Layer 3 escalations, and only with a specific indication-driven reason rather than an arc-driven one. The sequencing logic is identical to the one I described for taVNS; the difference is that taVNS can currently clear the bar for that slot and PBM cannot.

How I would parameterize it — and why I currently can't. Research-validated wavelength (commonly 810 nm, or 1064 nm in the cognitive work), measured irradiance, a target fluence chosen with the biphasic curve in mind, defined scalp sites, and — the part that makes this currently impractical — an honest estimate of the fraction of light actually reaching the target. The reason I cannot do this today is not that the parameters are unknowable in principle; it is that the affordable devices do not give me the irradiance transparency or the delivered-dose confidence I would need, and the devices that might are priced for a clinical evidence base that does not yet exist. Documentation discipline would be non-negotiable if I ever ran it — device, wavelength, irradiance, fluence, sites, duration, and a delivered-dose estimate every session — for the same reason it matters in EEG-NF and more so, because with a biphasic dose-response I could be missing the therapeutic window in either direction and never know.

How I would talk to the client about it. Honestly, and currently reserved: that it is a modality I find mechanistically interesting and clinically premature for brain targets, that the physics of getting light through the skull is a real and under-disclosed problem, and that I would rather not spend their money on an expensive maybe or a cheaper implausibility. If a client arrives already using a consumer red-light helmet, my job is not to sneer at it — it is to be straight about what the evidence does and does not support, to keep them safe (eye protection, thermal sense, no grandiose substitution for actual treatment), and to fold the honest uncertainty into the plan rather than either endorsing or dismissing the device wholesale.

So: three targets, framed as the conditions for a future yes rather than a present integration.

Watch the dosimetry disclosures. The single most useful thing the device market could do is publish per-area irradiance and honest delivered-dose estimates. When that starts happening — when a device tells me not just what it emits but what it plausibly delivers — the modality moves meaningfully closer to clinical defensibility. Until then, the black box is the answer.

Watch the controlled trials, especially the replications. One enthusiastic lab is a hypothesis. Replicated, sham-controlled, properly dosed trials in indications I treat are a clinical signal. NEST-3 is the standing reminder that even a well-funded pivotal program can fail, so I weight replication heavily and discount open-label enthusiasm.

Keep the NF-integration question open and honest. Whether transcranial PBM does anything to neurofeedback learning — applied around a session as a metabolic or perfusion prime — is a genuinely interesting question that almost no one has properly asked. I would support a careful pilot of it in an informed-client research context before I would ever offer it as a clinical service, and I would treat the result, positive or negative, as informative rather than threatening.

The integration verdict, then, is a clean no-for-now with a clearly specified path to yes. The high-yield move available today is not adoption; it is literacy — being the practitioner in the room who can explain, fairly and precisely, why the light helmet is interesting and not yet ready, and what would make it ready. That literacy protects clients from the cash grab and protects the field's credibility, and it costs nothing but the willingness to hold curiosity and skepticism in the same hand. I am writing this piece for the version of myself who will, I expect, eventually get to revisit the verdict — and for the reader who is being sold the revisited verdict today, years early.

Conclusion

Photobiomodulation is the modality in this series that I take seriously by watching it. The cellular mechanism is real — cytochrome c oxidase, the nitric-oxide brake-release, the metabolic nudge — and the peripheral evidence base for pain and healing is respectable. But the brain is behind a centimeter of scattering, absorbing bone, and most of the light never gets there; the transcranial clinical literature is young and heterogeneous; the one large pivotal brain trial was stopped for futility; and the version of the modality engineered to overcome the physics is expensive enough that the affordable alternatives slide toward selling people a pretty light. Those are not the conditions under which I integrate a tool into a practice I am accountable for.

That is not the same as dismissal, and I want the difference to be the thing you carry out of this piece. Sometimes taking a modality seriously means adopting it carefully, as it did with taVNS. Sometimes it means explaining, fairly and in detail, why you are not adopting it yet — and writing down the exact evidence and engineering that would change your mind. Skepticism that comes with a spec sheet for its own reversal is not closed-mindedness. It is the open-minded posture doing its actual job, which is not to believe everything that glows.

The light is real. The mechanism is real. The skull is also real, and so is the price tag, and so is the difference between a controlled trial's measured dose and a consumer helmet's hopeful one. Watch the dosimetry disclosures. Watch the replications. Keep the integration question open and honest. And when a device finally tells you not just what it emits but what it delivers, and a replicated trial in a real indication backs it up — that is the day this verdict gets revisited. Until then: curious, patient, and unbought.

References

• Cassano, P., Petrie, S. R., Mischoulon, D., Cusin, C., Katnani, H., Yeung, A., De Taboada, L., Archibald, A., Bui, E., Baer, L., Affı, A. M., Kim, J., Iuliano, J., Fava, M., & Iosifescu, D. V. (2018). Transcranial photobiomodulation for the treatment of major depressive disorder. The ELATED-2 pilot trial. Photomedicine and Laser Surgery, 36(12), 634–646. https://doi.org/10.1089/pho.2018.4490

• Hamblin, M. R. (2016). Shining light on the head: Photobiomodulation for brain disorders. BBA Clinical, 6, 113–124. https://doi.org/10.1016/j.bbacli.2016.09.002

• Hacke, W., Schellinger, P. D., Albers, G. W., Bornstein, N. M., Dahlof, B. L., Fulton, R., Kasner, S. E., Shuaib, A., Richieri, S. P., Dilly, S. G., Zivin, J., & Lees, K. R. (2014). Transcranial laser therapy in acute stroke treatment: Results of Neurothera Effectiveness and Safety Trial 3 (NEST-3), a phase III randomized clinical trial. Stroke, 45(11), 3187–3193. https://doi.org/10.1161/STROKEAHA.114.005795

• Huang, Y.-Y., Sharma, S. K., Carroll, J., & Hamblin, M. R. (2011). Biphasic dose response in low level light therapy — an update. Dose-Response, 9(4), 602–618. https://doi.org/10.2203/dose-response.11-009.Hamblin

• Jagdeo, J. R., Adams, L. E., Brody, N. I., & Siegel, D. M. (2012). Transcranial red and near infrared light transmission in a cadaveric model. PLoS ONE, 7(10), e47460. https://doi.org/10.1371/journal.pone.0047460

• Karu, T. I. (2008). Mitochondrial signaling in mammalian cells activated by red and near-IR radiation. Photochemistry and Photobiology, 84(5), 1091–1099. https://doi.org/10.1111/j.1751-1097.2008.00394.x

• Naeser, M. A., Zafonte, R., Krengel, M. H., Martin, P. I., Frazier, J., Hamblin, M. R., Knight, J. A., Meehan, W. P., & Baker, E. H. (2014). Significant improvements in cognitive performance post-transcranial, red/near-infrared light-emitting diode treatments in chronic, mild traumatic brain injury: Open-protocol study. Journal of Neurotrauma, 31(11), 1008–1017. https://doi.org/10.1089/neu.2013.3244

• Salehpour, F., Mahmoudi, J., Kamari, F., Sadigh-Eteghad, S., Rasta, S. H., & Hamblin, M. R. (2018). Brain photobiomodulation therapy: A narrative review. Molecular Neurobiology, 55(8), 6601–6636. https://doi.org/10.1007/s12035-017-0852-4

• Tedford, C. E., DeLapp, S., Jacques, S., & Anders, J. (2015). Quantitative analysis of transcranial and intraparenchymal light penetration in human cadaver brain tissue. Lasers in Surgery and Medicine, 47(4), 312–322. https://doi.org/10.1002/lsm.22343